Chinese AI Upstart DeepSeek Rattles Google With Model Matching Gemini’s Power

A newcomer in the artificial intelligence world is making waves across the global tech landscape. DeepSeek, a Chinese AI firm based in Hangzhou, has emerged as a surprising competitor to tech giants like Google and OpenAI just over a year after its founding in May 2023.

Founded by Liang Wenfeng, a Zhejiang University graduate, DeepSeek operates as an independent AI research lab under the quantitative hedge fund High-Flyer. The company has quickly gained international recognition after launching its first AI model in November 2023, followed by its groundbreaking R1 reasoning model in January 2025.

DeepSeek’s latest achievement, the DeepSeek-V3 model released in December 2024, performs at levels comparable to OpenAI’s GPT-4o and Anthropic’s Claude 3.5. These models can handle diverse tasks from answering questions to writing essays and generating computer code. While this advancement is impressive, the global AI market is projected to reach $1.5 trillion by 2030, indicating massive growth potential for companies like DeepSeek.

What’s particularly remarkable about DeepSeek’s success is its cost efficiency. The company developed its state-of-the-art models for under $6 million, compared to the hundreds of millions spent by U.S. companies. This achievement has disrupted industry expectations and even caused a 3% drop in Nvidia’s share price.

The company’s efficiency stems from innovative techniques based on “sparsity,” which selectively trains only relevant model parameters among hundreds of billions. DeepSeek has also developed advanced compression methods that store and access model information more efficiently.

The R1 model represents a significant breakthrough with its advanced reasoning capabilities. It can perform step-by-step problem-solving without requiring large supervised datasets, a feature considered the “holy grail” of AI development. The model utilizes a hybrid architecture training that enables both fast inference and deep reasoning capabilities.

DeepSeek’s rapid rise demonstrates a potential shift in global AI power dynamics, especially amid ongoing U.S.-China technology tensions. The company makes its models available through various interfaces including web platforms, mobile apps, and developer APIs, as well as on platforms like Hugging Face.

Industry experts now watch closely as this Chinese upstart continues to challenge the established players in the global AI race with its powerful, cost-effective models. The company’s decision to release its models under the MIT License allows researchers worldwide to freely access, modify, and use these advanced AI systems.